Degradation levels of continuous speech affect neural speech tracking and alpha power differently

TY - JOUR

T1 - Degradation levels of continuous speech affect neural speech tracking and alpha power differently

AU - Hauswald, Anne

AU - Keitel, Anne

AU - Chen, Ya-Ping

AU - Rösch, Sebastian

AU - Weisz, Nathan

N1 - Funding Information: This study was supported by a FWF Einzelprojekt (P 31230). AK was supported by the Wellcome Trust [204820/Z/16/Z]. We would like to thank Joachim Gross and Hyojin Park for providing their original MATLAB script for extracting the lip area. We would also like to thank Siri Ebert and Jonas Heilig for help with data acquisition. Publisher Copyright: © 2020 The Authors. European Journal of Neuroscience published by Federation of European Neuroscience Societies and John Wiley & Sons Ltd

PY - 2022/6/28

Y1 - 2022/6/28

N2 - Making sense of a poor auditory signal can pose a challenge. Previous attempts to quantify speech intelligibility in neural terms have usually focused on one of two measures, namely low-frequency speech-brain synchronization or alpha power modulations. However, reports have been mixed concerning the modulation of these measures, an issue aggravated by the fact that they have normally been studied separately. We present two MEG studies analyzing both measures. In study 1, participants listened to unimodal auditory speech with three different levels of degradation (original, 7-channel and 3-channel vocoding). Intelligibility declined with declining clarity, but speech was still intelligible to some extent even for the lowest clarity level (3-channel vocoding). Low-frequency (1–7 Hz) speech tracking suggested a U-shaped relationship with strongest effects for the medium-degraded speech (7-channel) in bilateral auditory and left frontal regions. To follow up on this finding, we implemented three additional vocoding levels (5-channel, 2-channel and 1-channel) in a second MEG study. Using this wider range of degradation, the speech-brain synchronization showed a similar pattern as in study 1, but further showed that when speech becomes unintelligible, synchronization declines again. The relationship differed for alpha power, which continued to decrease across vocoding levels reaching a floor effect for 5-channel vocoding. Predicting subjective intelligibility based on models either combining both measures or each measure alone showed superiority of the combined model. Our findings underline that speech tracking and alpha power are modified differently by the degree of degradation of continuous speech but together contribute to the subjective speech understanding.

AB - Making sense of a poor auditory signal can pose a challenge. Previous attempts to quantify speech intelligibility in neural terms have usually focused on one of two measures, namely low-frequency speech-brain synchronization or alpha power modulations. However, reports have been mixed concerning the modulation of these measures, an issue aggravated by the fact that they have normally been studied separately. We present two MEG studies analyzing both measures. In study 1, participants listened to unimodal auditory speech with three different levels of degradation (original, 7-channel and 3-channel vocoding). Intelligibility declined with declining clarity, but speech was still intelligible to some extent even for the lowest clarity level (3-channel vocoding). Low-frequency (1–7 Hz) speech tracking suggested a U-shaped relationship with strongest effects for the medium-degraded speech (7-channel) in bilateral auditory and left frontal regions. To follow up on this finding, we implemented three additional vocoding levels (5-channel, 2-channel and 1-channel) in a second MEG study. Using this wider range of degradation, the speech-brain synchronization showed a similar pattern as in study 1, but further showed that when speech becomes unintelligible, synchronization declines again. The relationship differed for alpha power, which continued to decrease across vocoding levels reaching a floor effect for 5-channel vocoding. Predicting subjective intelligibility based on models either combining both measures or each measure alone showed superiority of the combined model. Our findings underline that speech tracking and alpha power are modified differently by the degree of degradation of continuous speech but together contribute to the subjective speech understanding.

KW - MEG

KW - degraded speech

KW - low frequency speech tracking

KW - alpha power

KW - continuous speech

KW - low-frequency speech tracking

UR - https://www.scopus.com/pages/publications/85089071679

U2 - 10.1111/ejn.14912

DO - 10.1111/ejn.14912

M3 - Article

C2 - 32687616

SN - 0953-816X

VL - 55

SP - 3288

EP - 3302

JO - European Journal of Neuroscience

JF - European Journal of Neuroscience

IS - 11-12

ER -